Crypto Vision

A production-grade crypto intelligence platform — 200+ API endpoints, 37 data sources, 58 AI agents, real-time WebSocket feeds, anomaly detection, RAG-powered semantic search, BigQuery data warehouse, and a complete AI model training pipeline.

---

Table of Contents

• Overview
• Quick Start
• Environment Variables
• Architecture
• API Reference
• Data Sources
• Packages
• Apps
• AI Agents
• Workers & Ingestion
• Telegram Bot (crypto vision)
• Infrastructure
• Training & Fine-Tuning
• BigQuery Data Warehouse
• Testing
• Monitoring & Observability
• Self-Hosting
• Performance
• Project Structure
• Documentation

---

Overview

Crypto Vision is a comprehensive crypto data aggregation, intelligence, and AI platform built on TypeScript and the Hono framework. It provides:

• 200+ REST API endpoints across 35+ route domains (market data, DeFi, on-chain analytics, derivatives, news, AI, governance, macroeconomics, and more)
• 37 data source adapters (CoinGecko, DeFiLlama, Binance, Bybit, OKX, Deribit, dYdX, Hyperliquid, L2Beat, Messari, Token Terminal, GeckoTerminal, Jupiter, DePin Scan, etc.)
• 58 specialized AI agents for DeFi strategy, trading, portfolio management, security analysis, and education
• Real-time WebSocket feeds for prices, Bitcoin mempool, and DEX trades
• Anomaly detection engine using Modified Z-Score and EWMA algorithms across 16 anomaly types
• RAG-powered semantic search with multi-provider embeddings (Vertex AI, OpenAI)
• Multi-provider AI with automatic fallback chain: Groq → Gemini → OpenAI → Anthropic → OpenRouter
• Two-tier caching (in-memory LRU + Redis) with stampede protection and stale-while-revalidate
• Circuit breaker pattern per upstream host with exponential backoff
• BigQuery data warehouse with 22+ tables and materialized views
• Complete ML pipeline for model training, fine-tuning (Gemini + open-source), evaluation, and inference serving
• Full GCP infrastructure as code (Terraform, Cloud Run, Pub/Sub, Scheduler, Kubernetes)

---

Quick Start

Prerequisites

• Node.js ≥ 22 (ESM required)
• PostgreSQL 16 (for crypto vision)
• Redis 7 (optional, for distributed caching)
• Google Cloud credentials (optional, for BigQuery/Pub/Sub/Vertex AI)

Local Development

# Clone the repository
git clone https://github.com/nirholas/crypto-vision.git
cd crypto-vision

# Install dependencies
npm install

# Copy and configure environment variables
cp .env.example .env  # Edit with your API keys

# Start development server (hot-reload via tsx)
npm run dev

# Server starts on http://localhost:8080

Docker

# Build and run the main API server
docker build -t crypto-vision .
docker run -p 8080:8080 --env-file .env crypto-vision

# Or use Docker Compose for full stack (API + Redis)
docker compose up

Docker Compose — Ingestion Workers

# Start ingestion workers (market, defi, news, dex, derivatives, governance, macro, onchain)
docker compose -f docker-compose.ingest.yml up

NPM Scripts

Script	Description
npm run dev	Start dev server with hot-reload (tsx)
npm run build	Compile TypeScript (tsc -p tsconfig.build.json && tsc-alias)
npm start	Run compiled build (node dist/index.js)
npm test	Run all tests (Vitest)
npm run test:coverage	Run tests with coverage
npm run test:e2e	Run end-to-end tests
npm run lint	Run ESLint
npm run typecheck	TypeScript type checking
npm run db:generate	Generate Drizzle ORM migrations
npm run db:migrate	Apply database migrations
npm run db:push	Push schema changes
npm run db:studio	Open Drizzle Studio
npm run export:all	Export all data from BigQuery
npm run export:embeddings	Export embeddings
npm run import:postgres	Import exported data to PostgreSQL

---

Environment Variables

📖 See docs/SECURITY_GUIDE.md for secrets management best practices.

Core

Variable	Required	Default	Description
PORT	No	8080	Server port
NODE_ENV	No	development	Environment (development / production)
LOG_LEVEL	No	info	Pino log level

API Keys — Data Sources

Variable	Required	Description
COINGECKO_API_KEY	No	CoinGecko Pro/Demo API key
COINGECKO_API_TYPE	No	pro or demo (default: demo)
CRYPTOCOMPARE_API_KEY	No	CryptoCompare API key
ETHERSCAN_API_KEY	No	Etherscan API key
MESSARI_API_KEY	No	Messari API key
TOKENTERMINAL_API_KEY	No	Token Terminal API key
COINGLASS_API_KEY	No	CoinGlass API key
NEWSDATA_API_KEY	No	NewsData.io API key
CALENDARIFIC_API_KEY	No	Calendarific API key
GOPLUS_APP_KEY	No	GoPlus security API key
GOPLUS_APP_SECRET	No	GoPlus security API secret

AI Providers

Variable	Required	Description
GROQ_API_KEY	No	Groq API key (primary AI provider)
GEMINI_API_KEY	No	Google Gemini API key (fallback #1)
OPENAI_API_KEY	No	OpenAI API key (fallback #2)
ANTHROPIC_API_KEY	No	Anthropic API key (fallback #3)
OPENROUTER_API_KEY	No	OpenRouter API key (final fallback)

Database & Cache

Variable	Required	Description
DATABASE_URL	For Bot	PostgreSQL connection string
REDIS_URL	No	Redis connection string

Google Cloud

Variable	Required	Description
GCP_PROJECT_ID	For GCP features	Google Cloud project ID
GCP_BIGQUERY_DATASET	For BigQuery	BigQuery dataset name
GCP_PUBSUB_ENABLED	No	Enable Pub/Sub publishing (true/false)
VERTEX_AI_LOCATION	No	Vertex AI region (default: us-central1)

Authentication

Variable	Required	Description
API_KEY_PUBLIC	No	Public tier key (30 rpm)
API_KEY_BASIC	No	Basic tier key (200 rpm)
API_KEY_PRO	No	Pro tier key (2000 rpm)
API_KEY_ENTERPRISE	No	Enterprise tier key (10000 rpm)

Telegram Bot

Variable	Required	Description
SECTBOT_ENABLED	No	Enable crypto vision (true/false)
BOT_TOKEN	For Bot	Telegram bot token

---

Architecture

📖 See docs/ARCHITECTURE.md for the full architecture deep-dive.

┌──────────────────────────────────────────────────────────────────┐
│                        Clients                                   │
│   REST API  │  WebSocket  │  Telegram Bot  │  Dashboard/Apps     │
└──────┬──────┴──────┬──────┴───────┬────────┴──────┬──────────────┘
       │             │              │               │
┌──────▼─────────────▼──────────────▼───────────────▼──────────────┐
│                     Hono HTTP Server (src/index.ts)               │
│  ┌────────────────────────────────────────────────────────────┐   │
│  │ Middleware: CORS │ Auth │ Rate-Limit │ Compression │ ETag  │   │
│  │            │ Metrics │ Response Envelope │ Request Timeout │   │
│  └────────────────────────────────────────────────────────────┘   │
│                                                                   │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────────────┐ │
│  │ 35+ Route│  │ WebSocket│  │ Anomaly  │  │ AI + RAG Pipeline│ │
│  │ Modules  │  │ Manager  │  │ Engine   │  │ Multi-Provider   │ │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘  └───────┬──────────┘ │
└───────┼──────────────┼─────────────┼────────────────┼────────────┘
        │              │             │                │
┌───────▼──────────────▼─────────────▼────────────────▼────────────┐
│                        Core Libraries (src/lib/)                  │
│  ┌─────────┐  ┌────────────┐  ┌──────────┐  ┌────────────────┐  │
│  │ Fetcher │  │ Two-Tier   │  │ Bounded  │  │ Vector Store   │  │
│  │ Circuit │  │ Cache      │  │ Queue    │  │ (BigQuery /    │  │
│  │ Breaker │  │ LRU+Redis  │  │ AI:10/500│  │  In-Memory)    │  │
│  └────┬────┘  └─────┬──────┘  └────┬─────┘  └───────┬────────┘  │
└───────┼─────────────┼──────────────┼─────────────────┼───────────┘
        │             │              │                 │
┌───────▼─────────────▼──────────────▼─────────────────▼───────────┐
│               37 Data Source Adapters (src/sources/)              │
│  CoinGecko │ DeFiLlama │ Binance │ Bybit │ OKX │ L2Beat │ ...   │
└───────┬──────────────────────────────────────────────────────────┘
        │
┌───────▼──────────────────────────────────────────────────────────┐
│                    Storage & Messaging                            │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌────────────────┐   │
│  │PostgreSQL│  │  Redis   │  │ BigQuery │  │  Pub/Sub       │   │
│  │(Drizzle) │  │(ioredis) │  │ (22+ tbl)│  │ (5 tiers)     │   │
│  └──────────┘  └──────────┘  └──────────┘  └────────────────┘   │
└──────────────────────────────────────────────────────────────────┘

Key Architectural Patterns

Two-Tier Cache — In-memory LRU (200,000 entries) with optional Redis L2. Features stampede protection (coalesced requests), stale-while-revalidate (serve stale data while refreshing), per-key TTL, and background refresh.

Circuit Breaker — Per-host circuit breakers with configurable failure thresholds. States: CLOSED → OPEN (after 5 failures) → HALF-OPEN (probe after 30s). Prevents cascade failures to upstream APIs.

Bounded Concurrency Queue — AI requests are bounded to 10 concurrent / 500 max queued. Prevents overloading AI providers and manages backpressure.

Multi-Provider AI Fallback — Automatic failover chain: Groq → Gemini → OpenAI → Anthropic → OpenRouter. If one provider is down, the next in chain handles the request transparently.

4-Tier API Authentication — Public (30 rpm), Basic (200 rpm), Pro (2,000 rpm), Enterprise (10,000 rpm). Header-based API key authentication with per-tier rate limiting.

Response Envelope — All responses wrapped in { success, data, meta: { timestamp, source, cached, latency_ms } } for consistent client consumption.

---

API Reference

📖 See docs/API_REFERENCE.md for endpoint details and docs/API_AUTHENTICATION.md for auth tiers & rate limits.

The full OpenAPI 3.1 specification is available at openapi.yaml.

Health & Metrics

Method	Path	Description
GET	/health	Health check with cache stats, circuit breaker status, queue stats, WebSocket stats, memory usage
GET	/metrics	Prometheus metrics endpoint

Market Data (/api/market)

Method	Path	Description
GET	/api/market/prices	Current prices for top cryptocurrencies
GET	/api/market/prices/:id	Price data for a specific coin
GET	/api/market/trending	Trending coins
GET	/api/market/global	Global market statistics
GET	/api/market/dominance	Market dominance breakdown
GET	/api/market/fear-greed	Fear & Greed Index
GET	/api/market/history/:id	Historical price data
GET	/api/market/categories	Market categories
GET	/api/market/top-movers	Top gainers and losers
GET	/api/market/high-volume	High volume coins

DeFi (/api/defi)

Method	Path	Description
GET	/api/defi/protocols	DeFi protocol rankings
GET	/api/defi/tvl	Total Value Locked across DeFi
GET	/api/defi/yields	Yield farming opportunities
GET	/api/defi/stablecoins	Stablecoin analytics
GET	/api/defi/chains	Chain-level DeFi metrics
GET	/api/defi/dexes	DEX volume data
GET	/api/defi/bridges	Bridge volume and TVL
GET	/api/defi/liquidations	Liquidation data

News (/api/news)

Method	Path	Description
GET	/api/news	Latest crypto news
GET	/api/news/analysis	AI-analyzed news with sentiment

On-Chain (/api/onchain)

Method	Path	Description
GET	/api/onchain/gas	Gas prices (Ethereum)
GET	/api/onchain/whale-transactions	Whale transaction alerts
GET	/api/onchain/token-security/:address	Token security audit
GET	/api/onchain/contract-info/:address	Smart contract information

AI & Intelligence (/api/ai)

22 AI-powered endpoints for market analysis, portfolio advice, DeFi strategy, risk assessment, and more.

Method	Path	Description
POST	/api/ai/analyze	AI market analysis
POST	/api/ai/portfolio-advice	AI portfolio recommendations
POST	/api/ai/defi-strategy	DeFi strategy generation
POST	/api/ai/risk-assess	Risk assessment
POST	/api/ai/trading-signals	AI trading signal analysis
POST	/api/ai/sentiment	Sentiment analysis
POST	/api/ai/compare	Comparative token analysis
POST	/api/ai/predict	Price prediction analysis
POST	/api/ai/explain	Explain crypto concepts
POST	/api/ai/news-impact	News impact analysis
POST	/api/ai/whale-analysis	Whale behavior analysis
POST	/api/ai/yield-optimize	Yield optimization
POST	/api/ai/security-audit	Smart contract security audit
POST	/api/ai/gas-optimize	Gas optimization recommendations
POST	/api/ai/narrative	Market narrative detection
POST	/api/ai/correlation	Cross-asset correlation analysis
POST	/api/ai/tokenomics	Tokenomics analysis
POST	/api/ai/chain-compare	Multi-chain comparison
POST	/api/ai/defi-risk-score	DeFi protocol risk scoring
POST	/api/ai/portfolio-rebalance	Portfolio rebalancing suggestions
POST	/api/ai/entry-exit	Entry/exit point analysis
POST	/api/ai/macro-impact	Macro-economic impact on crypto

AI Agents (/api/agents)

Method	Path	Description
GET	/api/agents	List all 58 available agents
GET	/api/agents/:agentId	Get agent details
POST	/api/agents/:agentId/invoke	Invoke an agent with a query
POST	/api/agents/multi	Invoke multiple agents in parallel
POST	/api/agents/orchestrate	Multi-agent orchestration (planner → executor → synthesizer)
GET	/api/agents/categories	Agent categories
GET	/api/agents/search	Search agents

CEX Data (/api/cex)

Method	Path	Description
GET	/api/cex/orderbook/:symbol	Order book depth
GET	/api/cex/trades/:symbol	Recent trades
GET	/api/cex/ticker/:symbol	24h ticker
GET	/api/cex/klines/:symbol	Candlestick data
GET	/api/cex/funding/:symbol	Funding rates

DEX Data (/api/dex)

Method	Path	Description
GET	/api/dex/trending	Trending pools
GET	/api/dex/new-pairs	Newly created pairs
GET	/api/dex/pools/:network/:address	Pool details
GET	/api/dex/tokens/:network/:address	Token info from DEXes
GET	/api/dex/search	Search DEX pools

Derivatives (/api/derivatives)

Method	Path	Description
GET	/api/derivatives/options	Options market data
GET	/api/derivatives/futures	Futures market data
GET	/api/derivatives/open-interest	Open interest aggregation
GET	/api/derivatives/funding-rates	Cross-exchange funding rates
GET	/api/derivatives/liquidations	Liquidation data

Perps (/api/perps)

Method	Path	Description
GET	/api/perps/markets	Perpetual markets overview
GET	/api/perps/funding	Perpetual funding rates

Bitcoin (/api/bitcoin)

Method	Path	Description
GET	/api/bitcoin/stats	Bitcoin network statistics
GET	/api/bitcoin/mempool	Mempool status
GET	/api/bitcoin/hashrate	Hash rate data
GET	/api/bitcoin/fees	Fee estimates
GET	/api/bitcoin/lightning	Lightning Network stats

Gas (/api/gas)

Method	Path	Description
GET	/api/gas/ethereum	Ethereum gas prices
GET	/api/gas/multi-chain	Multi-chain gas prices
GET	/api/gas/history	Gas price history

Governance (/api/governance)

Method	Path	Description
GET	/api/governance/proposals	Active DAO proposals
GET	/api/governance/spaces	Snapshot spaces
GET	/api/governance/votes/:id	Votes for a proposal

Macro (/api/macro)

Method	Path	Description
GET	/api/macro/indicators	Macroeconomic indicators (from Yahoo Finance)
GET	/api/macro/correlation	Crypto-macro correlation

Solana (/api/solana)

Method	Path	Description
GET	/api/solana/tokens	Solana token data (via Jupiter)
GET	/api/solana/defi	Solana DeFi metrics

Additional Route Domains

Domain	Path Prefix	Description
DePIN	/api/depin	Decentralized Physical Infrastructure Networks
Exchanges	/api/exchanges	Exchange rankings and details
NFT	/api/nft	NFT collections and marketplaces
Whales	/api/whales	Whale wallet tracking
Staking	/api/staking	Staking yields and validators
Calendar	/api/calendar	Crypto events calendar
Oracles	/api/oracles	Oracle network data
Unlocks	/api/unlocks	Token unlock schedules
ETF	/api/etf	Crypto ETF data
Portfolio	/api/portfolio	Portfolio tracking and analysis
Social	/api/social	Social metrics and sentiment
L2	/api/l2	Layer 2 scaling data (L2Beat)
Security	/api/security	Token/contract security analysis (GoPlus)
Research	/api/research	Research reports and analysis
Analytics	/api/analytics	Platform analytics
Ecosystem	/api/ecosystem	Ecosystem mapping
Aggregate	/api/aggregate	Multi-source data aggregation
Export	/api/export	Data export (CSV, JSON, Parquet)
Search	/api/search	Semantic search across all data
Anomalies	/api/anomalies	Anomaly detection alerts
News Feed	/api/news-feed	Aggregated news feed (12+ sources)

WebSocket Endpoints

📖 See docs/WEBSOCKET.md for message schemas, throttling, and client examples.

Path	Description
/ws/prices	Real-time price stream (CoinCap, 5Hz throttle)
/ws/bitcoin	Bitcoin mempool & block updates (Mempool.space)
/ws/trades	Live DEX trades (DexScreener)
/ws/alerts	Anomaly detection alerts (real-time)

---

Data Sources

📖 See docs/DATA_SOURCES.md for adapter configuration and rate limit details.

37 adapters in src/sources/ connect to external APIs:

Source	File	Data
CoinGecko	coingecko.ts	Prices, market data, trending, categories, history
DeFiLlama	defillama.ts	TVL, yields, stablecoins, DEX volumes, bridges
Alternative.me	alternative.ts	Fear & Greed Index
Binance	binance.ts	Order books, trades, klines, ticker, funding
Bybit	bybit.ts	Order books, trades, klines, funding rates
OKX	okx.ts	Order books, trades, klines, funding rates
Deribit	deribit.ts	Options, futures, volatility surface
dYdX	dydx.ts	Perpetual markets, funding rates
Hyperliquid	hyperliquid.ts	Perpetual markets, funding, open interest
CoinCap	coincap.ts	Real-time prices (WebSocket upstream)
CoinGlass	coinglass.ts	Open interest, liquidations, funding
CoinLore	coinlore.ts	Global market stats
CryptoCompare	cryptocompare.ts	Social stats, historical data
GeckoTerminal	geckoterminal.ts	DEX pools, trending pairs, OHLCV
Jupiter	jupiter.ts	Solana token prices, DEX aggregation
L2Beat	l2beat.ts	Layer 2 TVL and risk data
Messari	messari.ts	Fundamental metrics, profiles
Token Terminal	tokenterminal.ts	Revenue, earnings, P/E ratios
Bitcoin/Blockchain	bitcoin.ts, blockchain.ts	Network stats, mempool, hashrate
EVM	evm.ts	EVM chain RPC data
GoPlus	goplus.ts	Token security audits
Snapshot	snapshot.ts	Governance proposals, voting
DePin Scan	depinscan.ts	DePIN project data
Yahoo Finance	macro.ts	Macro indicators (DXY, S&P 500, VIX, bonds)
NFT Sources	nft.ts	NFT collections and marketplaces
Oracles	oracles.ts	Chainlink, Band, Pyth data
Social	social.ts	Social metrics
Staking	staking.ts	Staking yields, validators
Unlocks	unlocks.ts	Token unlock schedules
Whales	whales.ts	Whale transaction tracking
Portfolio	portfolio.ts	Portfolio analytics
Calendar	calendar.ts, coinmarketcal.ts	Crypto events
News	crypto-news.ts, news-aggregator.ts	News from 12+ outlets
ETF	etf.ts	Crypto ETF flow data

---

Packages

📖 See docs/PACKAGES.md for full package API documentation.

Eight packages in packages/ provide standalone functionality:

@nirholas/erc8004-agent-runtime — packages/agent-runtime

ERC-8004 compliant agent runtime for decentralized AI agents with A2A (Agent-to-Agent) messaging and x402 micropayments.

Key exports:

• ERC8004Agent — Base agent class with lifecycle management
• A2AHandler — Agent-to-Agent communication protocol handler
• TaskManager — Task queue management for agent workloads
• x402Middleware — HTTP 402 micropayment middleware
• IdentityManager — On-chain identity management
• ReputationManager — Agent reputation scoring
• searchAgents, connectToAgent — Agent discovery and connection
• Middleware: authMiddleware, rateLimitMiddleware, loggingMiddleware

Stack: TypeScript, Hono, ethers.js, EventEmitter3

@nirholas/binance-mcp-server — packages/binance-mcp

MCP (Model Context Protocol) server for Binance exchange. Provides AI-accessible tools for spot trading, staking, wallet management, NFTs, Binance Pay, and mining.

Stack: TypeScript, @modelcontextprotocol/sdk, @binance/* SDK packages

@nirholas/bnbchain-mcp — packages/bnbchain-mcp

MCP server for BNB Chain with dual support for BSC (EVM smart contracts) and Greenfield (decentralized storage).

Stack: TypeScript, @modelcontextprotocol/sdk, @bnb-chain/greenfield-js-sdk, ethers.js

@nirholas/crypto-market-data — packages/market-data

Standalone crypto market data service, Edge Runtime compatible. Aggregates CoinGecko, DeFiLlama, and Fear & Greed data with built-in caching and rate limiting.

Stack: TypeScript, Edge Runtime compatible

@crypto-vision/mcp-server — packages/mcp-server

MCP server exposing Crypto Vision intelligence tools to AI models. Includes Solana integration.

Stack: TypeScript, @modelcontextprotocol/sdk, @solana/web3.js

sweep — packages/sweep

Multi-chain dust sweeper with DeFi routing. Sweeps small token balances across chains, consolidating them via DEX routes.

Stack: TypeScript, Hono, ethers.js, @solana/web3.js, bullmq, drizzle-orm

abi-to-mcp (UCAI) — packages/ucai

Universal Contract AI Interface — Python tool that generates MCP servers from Ethereum ABI files. Published on PyPI as abi-to-mcp. Web builder at mcp.ucai.tech.

Stack: Python, mcp, web3.py

📖 See docs/PACKAGES.md for full package API documentation.

Eight packages in packages/ provide standalone functionality:

@nirholas/erc8004-agent-runtime — packages/agent-runtime

ERC-8004 compliant agent runtime for decentralized AI agents with A2A (Agent-to-Agent) messaging and x402 micropayments.

Key exports:

• ERC8004Agent — Base agent class with lifecycle management
• A2AHandler — Agent-to-Agent communication protocol handler
• TaskManager — Task queue management for agent workloads
• x402Middleware — HTTP 402 micropayment middleware
• IdentityManager — On-chain identity management
• ReputationManager — Agent reputation scoring
• searchAgents, connectToAgent — Agent discovery and connection
• Middleware: authMiddleware, rateLimitMiddleware, loggingMiddleware

Stack: TypeScript, Hono, ethers.js, EventEmitter3

@nirholas/binance-mcp-server — packages/binance-mcp

MCP (Model Context Protocol) server for Binance exchange. Provides AI-accessible tools for spot trading, staking, wallet management, NFTs, Binance Pay, and mining.

Stack: TypeScript, @modelcontextprotocol/sdk, @binance/* SDK packages

@nirholas/bnbchain-mcp — packages/bnbchain-mcp

MCP server for BNB Chain with dual support for BSC (EVM smart contracts) and Greenfield (decentralized storage).

Stack: TypeScript, @modelcontextprotocol/sdk, @bnb-chain/greenfield-js-sdk, ethers.js

@nirholas/crypto-market-data — packages/market-data

Standalone crypto market data service, Edge Runtime compatible. Aggregates CoinGecko, DeFiLlama, and Fear & Greed data with built-in caching and rate limiting.

Stack: TypeScript, Edge Runtime compatible

@crypto-vision/mcp-server — packages/mcp-server

MCP server exposing Crypto Vision intelligence tools to AI models. Includes Solana integration.

Stack: TypeScript, @modelcontextprotocol/sdk, @solana/web3.js

Stack: TypeScript, Hono, ethers.js, grammy, bullmq, drizzle-orm

sweep — packages/sweep

Multi-chain dust sweeper with DeFi routing. Sweeps small token balances across chains, consolidating them via DEX routes.

Stack: TypeScript, Hono, ethers.js, @solana/web3.js, bullmq, drizzle-orm

abi-to-mcp (UCAI) — packages/ucai

Universal Contract AI Interface — Python tool that generates MCP servers from Ethereum ABI files. Published on PyPI as abi-to-mcp. Web builder at mcp.ucai.tech.

Stack: Python, mcp, web3.py

---

Apps

Three frontend applications in apps/:

Dashboard — apps/dashboard

crypto-data-aggregator — Next.js dashboard for real-time crypto market data visualization. Features DeFi analytics, portfolio tracking, and watchlists.

Stack: Next.js, React, TypeScript

News — apps/news

free-crypto-news — Next.js crypto news aggregator with AI-powered analysis. Aggregates 12+ sources with sentiment tracking. Includes MCP integration, mobile app, CLI tool, browser extension, and embeddable widget. README available in 50+ languages.

Stack: Next.js, React, TypeScript, MCP

Video — apps/video

@crypto-vision/video — Remotion project for generating x402 Gas Station promotional videos programmatically.

Stack: Remotion, React, TypeScript

---

AI Agents

📖 See docs/AI_AGENTS.md for the full agent catalog and prompt engineering guide.

58 specialized AI agents in the agents/ directory (@nirholas/ai-agents-library v1.42.0):

Agent Categories

Category	Count	Examples
Trading	8	Swing Trader, Position Sizer, MEV Shield, Leverage Calculator
DeFi	12	Yield Farmer, Protocol Comparator, Insurance Advisor, Risk Scoring Engine
Portfolio	5	Portfolio Optimizer, Tax Strategist, Dollar-Cost Averaging
Security	4	Bridge Security Analyst, Rug Pull Detector, Smart Contract Auditor
Education	6	DeFi Onboarding Mentor, APY vs APR Educator, Impermanent Loss Calculator
Research	8	Alpha Leak Detector, Whale Tracker, On-Chain Forensics, Governance Analyst
Ecosystem	15	Sperax-specific agents (SPA Staker, USDs Navigator, Demeter Farmer, etc.)

Agent Invocation

# List all agents
curl http://localhost:8080/api/agents

# Invoke a specific agent
curl -X POST http://localhost:8080/api/agents/defi-yield-farmer/invoke \
  -H "Content-Type: application/json" \
  -d '{"query": "Find the best stablecoin yields above 10% APY"}'

# Multi-agent orchestration
curl -X POST http://localhost:8080/api/agents/orchestrate \
  -H "Content-Type: application/json" \
  -d '{"query": "Analyze ETH risk and suggest a DeFi strategy", "agents": ["defi-risk-scoring-engine", "defi-yield-farmer"]}'

Supported Languages

Agent prompts are translated into 18 languages: English, Spanish, French, German, Portuguese, Italian, Russian, Chinese (Simplified & Traditional), Japanese, Korean, Arabic, Turkish, Hindi, Vietnamese, Thai, Indonesian, Dutch.

---

Workers & Ingestion

📖 See docs/WORKERS.md for worker configuration, scheduling, and troubleshooting.

15 worker files in src/workers/ provide background data ingestion:

Worker Architecture

All workers extend WorkerBase which provides:

• Periodic fetching with configurable intervals
• Dual-write to BigQuery (streaming inserts) and Pub/Sub
• Exponential backoff on failures
• Graceful shutdown handling
• Health reporting

Available Workers

Worker	File	Interval	Description
Market Ingest	ingest-market.ts	1-2 min	Prices, market cap, volume from CoinGecko
DeFi Ingest	ingest-defi.ts	5 min	TVL, yields, protocol data from DeFiLlama
News Ingest	ingest-news.ts	10 min	Crypto news from multiple outlets
DEX Ingest	ingest-dex.ts	2 min	DEX trades, pools, trending pairs
Derivatives Ingest	ingest-derivatives.ts	2 min	Options, futures, funding rates
Governance Ingest	ingest-governance.ts	30 min	DAO proposals, votes
Macro Ingest	ingest-macro.ts	1 hour	Macroeconomic indicators
On-Chain Ingest	ingest-onchain.ts	5 min	On-chain metrics, gas prices

Indexing Workers

Worker	File	Description
Agent Indexer	index-agents.ts	Indexes AI agents into vector store
Governance Indexer	index-governance.ts	Indexes governance proposals
News Indexer	index-news.ts	Indexes news articles for RAG
Protocol Indexer	index-protocols.ts	Indexes DeFi protocol data
Historical Backfill	backfill-historical.ts	Backfills historical data

Running Workers

# Via Docker Compose
docker compose -f docker-compose.ingest.yml up

# Individual workers
npm run worker:market
npm run worker:defi
npm run worker:news

# All workers
npm run workers

---

Telegram Bot (crypto vision)

📖 See docs/TELEGRAM_BOT.md for bot commands, configuration, and deployment.

A full-featured Telegram trading bot in src/bot/ built on the grammy framework.

Features

• Trading Calls — Post buy/sell calls with P&L tracking
• Leaderboard — Global and group leaderboards ranked by ROI
• Hardcore Mode — Gamified challenge mode with elimination and prizes
• Portfolio Tracking — Track positions across tokens
• PNL Cards — Visual profit/loss summary cards
• Insider Alerts — Smart money movement notifications
• Premium Tiers — Subscription-based premium features
• Referral System — User referral tracking and rewards
• Multi-Group Support — Works across multiple Telegram groups

Database Schema

PostgreSQL schema at src/bot/db/schema.ts (Drizzle ORM) includes tables for:

• Users, groups, channels
• Trading calls with entry/exit prices
• Leaderboard snapshots
• Hardcore mode sessions and participants
• PNL history
• Premium subscriptions
• Referral tracking
• Insider alerts

Enabling the Bot

# Set environment variables
SECTBOT_ENABLED=true
BOT_TOKEN=your-telegram-bot-token
DATABASE_URL=postgresql://user:pass@host:5432/sectbot

# The bot starts automatically with the main server
npm run dev

---

Infrastructure

📖 See docs/INFRASTRUCTURE.md for Terraform modules and docs/DEPLOYMENT.md for CI/CD pipelines.

Google Cloud Platform (GCP)

The primary deployment target is GCP Cloud Run with supporting services:

Service	Purpose
Cloud Run	Main API + worker containers
BigQuery	Data warehouse (22+ tables)
Pub/Sub	Event messaging (5-tier topic architecture)
Cloud Scheduler	Cron jobs for workers, retraining
Memorystore (Redis)	Distributed cache
Secret Manager	API keys and credentials
Vertex AI	Embeddings and model training
VPC	Private networking

Terraform

17 Terraform files in infra/terraform/ define the complete infrastructure:

cd infra/terraform
terraform init
terraform plan
terraform apply

Key files: main.tf, cloud_run.tf, redis.tf, bigquery.tf, pubsub.tf, scheduler.tf, secrets.tf, network.tf, iam.tf, monitoring.tf, vertex.tf, gke-gpu.tf, export.tf, apis.tf

Kubernetes

12 Kubernetes manifests in infra/k8s/ for GKE deployment:

• deployment.yaml — Main API deployment
• hpa.yaml — Horizontal Pod Autoscaler
• redis.yaml — Redis StatefulSet
• cronjobs.yaml — Ingestion worker CronJobs
• network-policy.yaml — Network policies
• training-job.yaml — ML training Job
• inference-deployment.yaml — Model inference serving

Pub/Sub Topic Architecture

5-tier messaging defined in infra/pubsub/topics.yaml:

Tier	Latency	Topics
Realtime	< 1s	market-prices-realtime, dex-trades-realtime
Frequent	1-2 min	market-data-frequent, dex-pools-frequent
Standard	5-10 min	defi-metrics-standard, news-articles-standard
Hourly	1 hour	governance-hourly, macro-indicators-hourly
Daily	24 hours	analytics-daily, embeddings-daily

CI/CD

Google Cloud Build pipelines:

• cloudbuild.yaml — Main API build and deploy
• cloudbuild-workers.yaml — Worker containers build and deploy

Quick Setup

# Full GCP setup (interactive)
bash infra/setup.sh

# Teardown
bash infra/teardown.sh

---

Training & Fine-Tuning

📖 See docs/TRAINING.md for the full training pipeline guide.

Complete ML pipeline in scripts/training/:

Training Data Generation

# Generate training data from BigQuery
npx tsx scripts/training/generate-training-data.ts

# Validate generated data
npx tsx scripts/training/validate-data.ts

Gemini Fine-Tuning

# Fine-tune Gemini via Vertex AI
npx tsx scripts/training/finetune-gemini.ts

# Evaluate fine-tuned models
npx tsx scripts/training/eval-models.ts

Open-Source Model Training

# Prepare data in Alpaca format
npx tsx scripts/training/opensource/prepare-data.ts

# Train with LoRA (QLoRA 4-bit quantization)
python scripts/training/opensource/train.py

# Benchmark against baselines
python scripts/training/opensource/benchmark.py

# Export to GGUF/ONNX
python scripts/training/opensource/export.py

Inference Serving

# Serve fine-tuned model with vLLM
python scripts/inference/serve.py

# Health check
python scripts/inference/healthcheck.py

Kubernetes resources: infra/k8s/training-job.yaml, infra/k8s/inference-deployment.yaml

---

BigQuery Data Warehouse

📖 See docs/BIGQUERY.md for schema details, partitioning, and query examples.

22+ tables defined in infra/bigquery/tables.sql:

Table Categories

Category	Tables
Market	market_prices, market_global, fear_greed_index
DeFi	defi_protocols, defi_yields, defi_tvl_chains, defi_stablecoins, dex_pools
Derivatives	derivatives_options, derivatives_futures, derivatives_funding, derivatives_oi
Blockchain	onchain_gas, bitcoin_stats, bitcoin_mempool
Governance	governance_proposals, governance_votes
News	news_articles
Macro	macro_indicators
AI	embeddings, search_queries, ai_completions

Materialized Views

Defined in infra/bigquery/views.sql:

• Hourly/daily price aggregations
• Protocol TVL rankings
• Cross-exchange funding rate comparisons
• Market anomaly summaries

Data Export

# Export all BigQuery data
npm run export:all

# Export embeddings
npm run export:embeddings

# Import to PostgreSQL
npm run import:postgres

---

Testing

📖 See docs/TESTING.md for test writing guidelines and CI integration.

Comprehensive test suite using Vitest:

Test Categories

Type	Location	Count	Description
Unit	tests/lib/	33 files	Every core library module
Route	tests/routes/	10 files	API route handlers
Integration	tests/integration/	1 file	End-to-end API flows
E2E	tests/e2e/	1 file	Smoke tests against running server
Benchmarks	tests/benchmarks/	1 file	Performance benchmarks
Fuzz	tests/fuzz/	1 file	Fuzz testing for API inputs
Load	tests/load/	3 files	Load tests (smoke, soak, stress)

Running Tests

# All tests
npm test

# With coverage
npm run test:coverage

# E2E tests (requires running server)
npm run test:e2e

# Specific test file
npx vitest tests/lib/cache.test.ts

# Watch mode
npx vitest --watch

Key Test Files

• tests/lib/cache.test.ts — Two-tier cache with stampede protection
• tests/lib/fetcher.test.ts — Circuit breaker and retry logic
• tests/lib/anomaly.test.ts — Anomaly detection algorithms
• tests/lib/queue.test.ts — Bounded concurrency queue
• tests/lib/auth.test.ts — API key authentication tiers
• tests/lib/search.test.ts — Semantic search and intent detection
• tests/lib/embeddings.test.ts — Multi-provider embedding generation
• tests/lib/rag.test.ts — RAG pipeline
• tests/lib/orchestrator.test.ts — Multi-agent orchestration

---

Monitoring & Observability

📖 See docs/MONITORING.md for Grafana dashboards, alerting rules, and docs/ANOMALY_DETECTION.md for the anomaly detection engine.

Prometheus Metrics

Available at /metrics:

Metric	Type	Description
http_requests_total	Counter	Total HTTP requests by method, path, status
http_request_duration_seconds	Histogram	Request latency distribution
upstream_requests_total	Counter	Upstream API calls by source, status
upstream_request_duration_seconds	Histogram	Upstream latency
cache_hits_total / cache_misses_total	Counter	Cache hit/miss rates
cache_size	Gauge	Current cache size
websocket_connections	Gauge	Active WebSocket connections
queue_size	Gauge	Current queue depth
queue_concurrency	Gauge	Active concurrent tasks
circuit_breaker_state	Gauge	Circuit breaker states per host

Structured Logging

Pino-based structured JSON logging with:

• Request ID tracking
• Latency measurement
• Error context
• Configurable log levels via LOG_LEVEL

Health Check

GET /health returns:

{
  "status": "healthy",
  "uptime": 3600,
  "cache": { "size": 1500, "hitRate": 0.85 },
  "circuitBreakers": { "coingecko": "CLOSED", "defillama": "CLOSED" },
  "queue": { "pending": 2, "active": 1 },
  "websocket": { "connections": 15 },
  "memory": { "heapUsed": "150MB", "rss": "250MB" }
}

---

Self-Hosting

See docs/SELF_HOSTING.md for the complete self-hosting guide, including:

• Minimal setup (Node.js only, no cloud dependencies)
• Docker deployment
• Full GCP deployment with Terraform
• Scaling and production hardening
• Cost estimation

---

Performance

See docs/PERFORMANCE.md for detailed benchmarks and optimization guide, including:

• Response time targets (< 200ms cached, < 2s upstream)
• Cache optimization strategies
• Connection pooling
• Memory management
• Load testing methodology

---

Project Structure

crypto-vision/
├── src/                        # Main application source
│   ├── index.ts                # Entry point — Hono server, middleware, route mounting
│   ├── routes/                 # 40 route modules (market, defi, ai, agents, etc.)
│   ├── sources/                # 37 data source adapters
│   ├── lib/                    # Core libraries (cache, fetcher, queue, auth, anomaly, etc.)
│   ├── workers/                # 15 background ingestion & indexing workers
│   └── bot/                    # Telegram crypto vision (grammy)
│       └── db/schema.ts        # PostgreSQL schema (Drizzle ORM)
├── packages/                   # Standalone packages
│   ├── agent-runtime/          # ERC-8004 agent runtime (A2A + x402)
│   ├── binance-mcp/            # Binance MCP server
│   ├── bnbchain-mcp/           # BNB Chain MCP server
│   ├── market-data/            # Standalone market data service
│   ├── mcp-server/             # Crypto Vision MCP server
│   ├── pump-agent-swarm/       # Pump.fun agent swarm
│   ├── sweep/                  # Multi-chain dust sweeper
│   └── ucai/                   # ABI-to-MCP generator (Python)
├── apps/                       # Frontend applications
│   ├── dashboard/              # Next.js market dashboard
│   ├── news/                   # Next.js crypto news aggregator
│   └── video/                  # Remotion video project
├── agents/                     # 58 AI agents library
│   ├── prompts/                # Agent system prompts
│   ├── locales/                # 18 language translations
│   ├── schema/                 # Agent JSON schemas
│   └── docs/                   # Agent documentation
├── infra/                      # Infrastructure as Code
│   ├── terraform/              # 17 Terraform configs (GCP)
│   ├── k8s/                    # 12 Kubernetes manifests
│   ├── pubsub/                 # Pub/Sub topic definitions
│   ├── bigquery/               # BigQuery schema & views
│   └── scheduler/              # Cloud Scheduler jobs
├── scripts/                    # Operational scripts
│   ├── training/               # ML training pipeline
│   ├── inference/              # Model serving
│   ├── demo/                   # Demo scripts
│   ├── export-all.ts           # BigQuery data export
│   └── import-to-postgres.ts   # PostgreSQL import
├── tests/                      # Comprehensive test suite
│   ├── lib/                    # 33 unit test files
│   ├── routes/                 # 10 route test files
│   ├── e2e/                    # End-to-end smoke tests
│   ├── integration/            # API integration tests
│   ├── benchmarks/             # Performance benchmarks
│   ├── fuzz/                   # Fuzz testing
│   └── load/                   # Load tests (smoke, soak, stress)
├── docs/                       # Documentation
│   ├── AGENTS.md               # 58 AI agents guide
│   ├── API_REFERENCE.md        # Complete 300+ endpoint reference
│   ├── ARCHITECTURE.md         # System architecture & data flow
│   ├── COINGECKO_RATE_LIMITING.md # CoinGecko rate limiting strategy
│   ├── CONFIGURATION.md        # Environment variables reference
│   ├── DATABASE.md             # PostgreSQL & BigQuery schemas
│   ├── DATA_PIPELINE.md        # Workers, ingestion, Pub/Sub
│   ├── DATA_SOURCES.md         # 37+ upstream data sources
│   ├── DEPLOYMENT.md           # Docker, Cloud Run, K8s deployment
│   ├── DEVELOPER_WORKFLOW.md   # Day-to-day dev commands
│   ├── INFRASTRUCTURE.md       # Terraform, K8s, CI/CD
│   ├── ML_TRAINING.md          # Model training & fine-tuning
│   ├── PACKAGES.md             # All 8 packages deep-dive
│   ├── PERFORMANCE.md          # Performance & optimization
│   ├── REPOSITORY_GUIDE.md     # Full repo structure
│   ├── SELF_HOSTING.md         # Self-hosting guide
│   ├── TELEGRAM_BOT.md         # crypto vision documentation
│   ├── TESTING.md              # Test strategy & coverage
│   └── X402_PAYMENTS.md        # x402 micropayment system
├── prompts/                    # Engineering prompt guides
├── docker-compose.yml          # API + Redis stack
├── docker-compose.ingest.yml   # Ingestion workers
├── Dockerfile                  # Main API container
├── Dockerfile.worker           # Worker container
├── Dockerfile.train            # Training container
├── cloudbuild.yaml             # CI/CD — API
├── cloudbuild-workers.yaml     # CI/CD — Workers
├── drizzle.config.ts           # Drizzle ORM config
├── openapi.yaml                # OpenAPI 3.1 specification
├── vitest.config.ts            # Test configuration
└── tsconfig.json               # TypeScript configuration

---

Documentation

Comprehensive documentation is available in the docs/ directory:

Document	Description
AGENTS.md	58 AI agents — categories, architecture, orchestration, localization
ANOMALY_DETECTION.md	Real-time statistical anomaly detection — 16 event types, algorithms, processors
API_AUTHENTICATION.md	4-tier API key authentication, rate limiting, key management API
API_REFERENCE.md	Complete 538+ endpoint reference across 40 route modules
ARCHITECTURE.md	System architecture, data flow, caching, circuit breakers
COINGECKO_RATE_LIMITING.md	CoinGecko API rate limiting strategy and optimization
CONFIGURATION.md	All environment variables, API keys, secrets management
DATABASE.md	PostgreSQL schema (Drizzle ORM) and BigQuery table definitions
DATA_PIPELINE.md	Ingestion workers, Pub/Sub tiers, scheduling
DATA_SOURCES.md	37+ upstream data source adapters and their capabilities
DEPLOYMENT.md	Docker, Cloud Run, Kubernetes deployment guides
DEVELOPER_WORKFLOW.md	Day-to-day development commands and workflows
INFRASTRUCTURE.md	Terraform, GCP resources, K8s manifests, CI/CD pipelines
ML_TRAINING.md	Model fine-tuning (Gemini, Llama, Mistral), GPU training, vLLM inference
MONITORING.md	Prometheus metrics, structured logging, health checks, alerting
PACKAGES.md	All 8 packages — agent-runtime, MCP servers, market-data, sweep, ucai
PERFORMANCE.md	Benchmarks, optimization techniques, caching strategies
REPOSITORY_GUIDE.md	Full repository structure walkthrough
SECURITY_GUIDE.md	Security architecture, threat mitigations, best practices
SELF_HOSTING.md	Complete self-hosting guide with data migration
TELEGRAM_BOT.md	crypto vision — Telegram bot commands, features, database
TESTING.md	Test strategy, coverage, unit/integration/e2e/fuzz/load tests
TROUBLESHOOTING.md	Common issues, diagnostics, and solutions
WEBSOCKET.md	Real-time WebSocket feeds — prices, Bitcoin, trades, alerts
X402_PAYMENTS.md	x402 micropayment protocol integration

---

License

AGPL-3.0

---

Contributing

See CONTRIBUTING.md for contribution guidelines and SECURITY.md for security policy.

---

Built with TypeScript, Hono, and an unhealthy obsession with crypto data.